In a perfect world, organizations that transact business on the Internet in a global environment would translate all of their content into every language that their customers use. Research demonstrates that customers are dramatically more likely to consummate a transaction if the content is presented in their native language. The reality is that human translation is expensive and time consuming. Even the largest organizations with significant translation budgets are only able to translate a small fraction of their content using traditional human translation methods.
Increasingly, organizations are turning to machine translation (MT) solutions that automatically translate text segments, which may be words, phrases, or sentences, depending on the MT engine's capabilities, from one language to another. The quality of MT is nearly always inferior to human translation, however, the cost of MT is a fraction (in some cases, 1/10000th) of the cost of human translation and the results are often nearly instantaneous.
A specialized application of MT is the ability to translate content in real-time. Real-time MT provides the ability to translate dynamically-generated content. Some websites contain dynamically-generated content, such as news-feeds or user-controlled content (e.g. job postings, items for sale, etc.). An instant message or a transcript of a political debate are further examples of dynamically-generated content and for which immediate translation may be desirable.
MT can be implemented using either an on-premise solution or via software as a service (SaaS). On-premise solutions are implemented using servers that are typically co-located with the content to be translated and are typically operated by the content providers. SaaS solutions are multi-tenant platforms where many different customers are serviced by a single implementation typically provided by a third party over the Internet. Although, in SaaS, a single software executable may be used by all, each customer is usually allocated their own private “tenancy” for their data, which may be secured. Salesforce.com is a prototypical SaaS application.
Some of the better-known MT engines include: GOOGLE Translate (google.com/translate), MICROSOFT Translator (microsofttranslator.com), PROMT (promt.com), SYSTRAN (systransoft.com), and IBM n.Fluent (vvww.research.ibm.com/social/projects_nfluent.html).
While the underlying technologies of existing MT engines vary, their concepts are fundamentally the same: A user submits a translation request that contains information about the source material by either explicitly including the source text or by supplying a Uniform Resource Locator (URL) for a source document. The user further submits explicit information about the translation task, such as an identification of the target language and possibly parameters that define one or more translation options. Some engines require the user to also identify the source language whereas others can automatically detect the original language from the submitted source material. The MT engine parses the request, performs the requested task, and returns the translated content. Some MT engines are able to parse submissions and extract segments in order to translate the entire submission, whereas others require separate submission of each segment.
Real-time, SaaS, MT engines typically communicate with the requesting application via an application programming interface (API). Using an API, commercial developers can implement MT solutions that are integrated into existing applications as well as into new software applications (e.g., custom software applications developed to solve particular business problems known as purpose-built web sites). While APIs can be implemented in a variety of ways, many use Extensible Markup Language (XML) to exchange content between the client (the requesting application system) and the MT server.
Existing MT solutions have been used to “interactively” translate a given web page (for example, a Chinese website may be translated into English by the user clicking on a button for translation), and are known to be useful for looking up a given phrase, or to translate a document or email.
Existing MT solutions are limited because they either require the end user to visit a particular website and input the data to be translated or they require the developer to add code (e.g., JavaScript) to the webpage allowing translation. The necessity of having developers include code to direct the translation can be inconvenient. For example, including such code can be cumbersome in organizations where the content of the website is controlled by one group (e.g., Marketing) and the hosting operations are controlled by another (e.g., the IT department, or even a third party). The inconvenience is especially apparent and difficult to overcome where a content management system (CMS) is used that is incompatible with the translation APIs. Moreover, translated websites where code is used to interactively translate the website cannot be indexed by search engines. A further limitation to the interactive translation approach with an SaaS MT solution is that the content is transmitted over the public Internet, making this approach less desirable for sensitive intranet websites, for one example.
There is a need for methods and systems that facilitate website translation where translation code does not need to be added to the website. Solutions that facilitate indexing of translated content by search engine crawlers are also desirable. Also desirable is a secure solution appropriate for translating proprietary, confidential, or sensitive enterprise content with an SaaS MT engine.